Last spring [2023], in our NNAI faculty seminar, we listened to a presentation given to the US President’s Council of Advisors on Science and Technology by the economist Darin Acemoglu, who has become something of an expert commentator on AI.
I wanted to share some sections of the text of the talk that I thought were particularly interesting, so I asked an AI transcription service to provide a transcript of the audio. As many of you know, the tools for transcribing human voice have gotten quite excellent in a short period of time, even just the last 2 or 3 years.
I was thus provided with a nearly perfect transcription of Dr. Acemoglu’s peach, even given his relatively heavy Greek accent in English.
Nearly perfect. The one word that the large language model which translated the text could not seem to make sense of was the acronym “LLM” Large Language Model. Instead, in every instance where Acemoglu said LLM, it transcribed the word “lamb“. Thus:
“Automation is, of course, part of what we are going to do with machines. But it’s not the only thing we can do with machines. It’s not the only thing we can do with digital technologies or lambs.”
“Lambs remain illegible, meaning that what they do is not easy to understand for humans.”
“I argue that this sort of excessive, authoritative, illegible approach of lambs is really a major roadblock.”
“I think it’s easy to understand as lambs start populating more of the Internet, meaning that more posts on social media, more articles, more social media content is generated, at least with input from lambs.”
“I think the danger here is really quite present and real as lambs become more widespread.”
I could spin a deep story about the connections: about animal cognition, the much maligned “sheepish” behavior or a creature that is actually very intelligent or the role that domestication plays in producing intelligence or the diverse roles that the lamb plays in the literary imagination… but in reality the response to the lamb is
more visceral, it’s a kind of fellowship that is hard to put your finger on.
I could also use the ironic inability of a large language model to understand a word that refers to itself to spin a story of the impossibility of general artificial intelligence, a story of schadenfreude in the face of engineering hubris. But I don’t think that’s what most of us really believe, however funny it is to hear about the strange and powerful things lambs are capable of.
Join Our Newsletter
Connect
UCLA Institute for Society and Genetics
621 Charles E. Young Dr. South
Box 957221, 3360 LSB
Los Angeles, CA 90095-7221
Contact
"*" indicates required fields